Performance-Based Earthquake Engineering 2.0: Machine-Learning and Artificial Intelligence Algorithms for seismic hazard and vulnerability.

这一博士研究项目将极大地推动基于性能的地震工程(PBEE)领域的最先进革命，生产出健壮、快速、灵活和智能的新工具。该项目属于EPSRC工程研究领域，重点是如何将人工智能(EPSRC Area1)整合到土木工程中，特别强调地震工程、结构工程(EPSRC区域2)和地面工程(EPSRC区域3)。背景：全球每年发生150次6级以上地震，过去十年的灾难性事件突显了它们可能造成的生命损失和经济损失。尽管地震事件具有重要的社会和经济意义，但当代对地震引起的地面运动的预测仍然是基于经验模型，通过对历史观测进行统计回归，而几乎没有来自基本物理的输入。地震工程问题的几个方面还需要有能力的数据处理/数据挖掘工具来理解全球各地各种地面运动测量仪器记录的海量数据集。机器学习(ML)领域在过去几年中迅速发展，承诺修改数据科学在不同学科中的角色。ML在处理复杂问题、提高计算效率、传播和处理不确定性以及促进决策方面具有优势。随着ML的进步，它的应用不仅限于人工智能(AI)领域，而且已经扩展到科学和工程的几个领域。ML已经成为一种很有前途的数据推进工具，可以处理创建的海量数据集，并解决地震工程中各种具有挑战性的问题。机器学习在地震工程中的一些应用已经被一些研究人员付诸实践。相比之下，许多应用，如地面运动的实时预测和地震的破坏检测，主要是未开发的。因此，加速它们在地震工程领域的应用仍然存在重大机会。因此，需要对ML和地震工程的交叉领域进行更多的研究，以拥抱下一代数据共享和传感器技术，实现更先进的ML技术，并开发物理制导的ML模型。今后，本研究致力于了解在地震工程问题中实施ML的未来。研究目标和结果：本研究旨在将新兴的ML工具与经典的地震工程方法统一和协调，以创建一种基于性能的现代地震工程设计和评估方法。在这项研究中，将探索各种ML应用/算法。目的和目标：这项研究的一些预期结果包括：地震危险性分析。通过稀疏观测和模拟的综合来实时预测地面运动。地震易损性评估。图像检测，以开发基础设施漏洞的下一代数据集。用于减震的结构控制。基于高性能计算的模拟的元建模。识别和损坏检测。预测地震期间地面崩塌的预期移动。

This PhD research project would significantly advance the state-of-the-art revolutionising the field of performance-based earthquake engineering (PBEE), producing new tools that are robust, fast, flexible, and smart. This project falls within the EPSRC Engineering research area, focusing on how artificial intelligence (EPSRC Area1) can be integrated into civil engineering with a specific emphasis on earthquake engineering, structural engineering (EPSRC Area 2) and Ground Engineering (EPSRC Area 3). Context: Globally, 150 earthquakes over magnitude six are experienced yearly, and catastrophic events in the past decade have emphasised the loss of life and economic damage they can produce. Despite the social and economic importance of seismic events, contemporary prediction of earthquake-induced ground motions is still based on empirical models via statistical regression of historical observations with little input from the underlying physics. Several aspects of Earthquake Engineering problems also require capable data-handling/data-mining tools to understand huge volumes of datasets recorded by various ground motions measuring instruments placed across the globe. The field of Machine Learning (ML) has rapidly evolved over the past few years with the promise to modify the role of data science in various disciplines. ML offers advantages in handling complex problems, providing computational efficiency, propagating and treating uncertainties, and facilitating decision-making. With the advancement of ML, its application is not only limited to the field of Artificial Intelligence (AI) but has expanded to several fields of science and engineering. ML has emerged as a promising data-advancement tool to handle vast volumes of datasets created and solve various challenging problems in earthquake engineering. Some applications of machine learning in Earthquake Engineering have already been in practice by several researchers. In contrast, many applications, such as real-time prediction of ground motions and damage detection of earthquakes, have been mainly untapped. Hence, significant opportunities still exist to accelerate their applications in this field of Earthquake Engineering. Therefore, more research on the cross-field of ML and earthquake engineering is needed to embrace the next generation of data sharing and sensor technologies, implement more advanced ML techniques, and develop physics-guided ML models. Henceforth, this research is directed towards understanding the future of implementing ML into Earthquake Engineering problems. Research Objectives & Outcomes: This research aims to unify and harmonise the emerging ML tools with classical Earthquake Engineering approaches to create a modern Performance-Based Earthquake Engineering design and assessment approach. Various ML applications/algorithms will be explored in this research study. Aims and Objectives: Some of the expected outcomes from this study include: Seismic Hazard Analysis. Real-time prediction of ground motion via synthesis of sparse observations and simulations. Seismic fragility assessment. Image Detection to develop next-generation datasets of infrastructure vulnerability. Structural control for earthquake mitigation. Meta-modelling with High-Performance-Computing-based simulation. Identification and Damage Detection. Predicting the movement expected from ground failure during an earthquake.